RIT on Statistics Archives for Academic Year 2014


Organizational Meeting

When: Tue, September 10, 2013 - 3:30pm
Where: Math 1308
Speaker: Yuan Liao, UMCP

Intro to Penalized Least Squares

When: Tue, September 17, 2013 - 3:30pm
Where: MATH B0421
Speaker: Dr. Yuan Liao (Dept. of Math, UMCP) -
Abstract: We will start to discuss linear regression model with many regressors. The penalized least squares will be introduced, with a focus on the l_1 penalty, often known as ``Lasso". Intuitions and algorithms are to be discussed. If time permits, statistical properties will also be introduced.

Penalized Least Squares (Cont'd)

When: Tue, October 1, 2013 - 3:30pm
Where: MATH B0421
Speaker: Dr. Yuan Liao (UMCP) -

On the number of population moments simultaneously consistently estimable as a function of the (large) sample size

When: Tue, October 15, 2013 - 3:30pm
Where: MTH B0421
Speaker: Prof. Abram Kagan (Department of Mathematics, UMCP) -

Variable selection using L_1 penalized regression

When: Tue, October 22, 2013 - 3:30pm
Where: MTH 1313
Speaker: Xia Li and Yuan Liao, UMCP () -
Abstract: We shall see that LASSO is not variable selection consistent in general when the important and unimportant regressors are correlated. Intuitively, it puts equal weights to all the coefficients. A more "adaptive" penalty should penalize coefficients unequally. We shall discuss weighted L_1 penalized regression and its "oracle properties".

LARS for LASSO

When: Tue, October 29, 2013 - 3:30pm
Where: MTH 1313
Speaker: David Shaw (UMCP) -

Oracle Properties and Concave Penalties

When: Tue, November 5, 2013 - 3:30pm
Where: MTH B0421
Speaker: Prof. Paul Smith (Department of Mathematics, UMCP) -

Oracle Properties and Concave Penalties

When: Tue, November 12, 2013 - 3:30pm
Where: MATH B0421
Speaker: Paul Smith (UMCP) -

One-step Sparse Estimates in Nonconcave Penalized Likelihood Models

When: Tue, November 19, 2013 - 3:30pm
Where: MTH 1313
Speaker: Yue Tian (UMCP) -

Regularization and Variable Selection via the Elastic Net

When: Tue, November 26, 2013 - 3:30pm
Where: MTH 1313
Speaker: Hechao Sun (Department of Mathematics, UMCP) -
Abstract: This paper is written by Zou and Hastie (2005), with 2327 citations on Google scholar. They propose the elastic net, a new regularization and variable selection method. Real world data and a simulation study show that the elastic net often outperforms the lasso, while enjoying a similar sparsity of representation. In addition, the elastic net encourages a grouping effect, where strongly correlated predictors tend to be in or out of the model together.The elastic net is particularly useful when the number of predictors (p) is much bigger than the number of observations (n). By contrast, the lasso is not a very satisfactory variable selection method in the pn case. An algorithm called LARS-EN is proposed for computing elastic net regularization paths ef´Čüciently, much like algorithm LARS does for the lasso.

Plans for the Spring 2014 Semester

When: Tue, December 3, 2013 - 3:30pm
Where: B0421
Speaker: Paul Smith, Abram Kagan (UMCP) -

TBA

When: Tue, February 4, 2014 - 3:30pm
Where: MTH 1313
Speaker: () -

Nonparametric Estimation: An Introduction

When: Tue, February 25, 2014 - 3:30pm
Where: MTH 1313
Speaker: Prof. Abram Kagan (Department of Mathematics, UMCP) -

Nonparametric estimation: an introduction (Cont'd)

When: Tue, March 4, 2014 - 3:30pm
Where: MTH 1313
Speaker: Prof. Abram Kagan (Dept. of Statistics, UMCP) -

Necessary and sufficient condition for L1-consistency of the kernel estimators of density

When: Tue, March 11, 2014 - 3:30pm
Where: MTH 1313
Speaker: Prof. Abram Kagan (UMCP) -

Least Squares Cross-Validation in Density Estimation

When: Tue, March 25, 2014 - 3:30pm
Where: MTH 1313
Speaker: Hechao Sun (Dept. of Math, UMCP) -

The Hoeffding Inequality for Sum of Random Variables

When: Tue, April 1, 2014 - 3:30pm
Where: MTH 1313
Speaker: Xia Li (Dept. of Math., Univ. of Maryland) -

TBA

When: Tue, April 8, 2014 - 3:30pm
Where: MTH 1313
Speaker: David Shaw (Dept. of Math, UMCP) -

Generalized Additive Models

When: Tue, April 29, 2014 - 3:30pm
Where: MATH 1313
Speaker: Prof. Paul Smith (Dept. of Math., Univ. of Maryland) -

Inference from samples from location-scale parameters distributions

When: Tue, May 6, 2014 - 3:30pm
Where: MATH 1313
Speaker: Prof. Abram Kagan (Dept. of Math, Univ. of Maryland) -